IIIT Hyderabad researchers develop adaptive drones for flood rescue and agriculture
HYDERABAD: When a drone drops a payload onto a moving platform during a flood, it rapidly ascends because it suddenly loses weight. This abrupt change in flight dynamics challenges most engineers. Prof. Spandan Roy and his team at the Robotics Research Center of IIIT Hyderabad (IIITH) focus their work on addressing this challenge.
“Even if you don’t know the system, can you still control it?” This unconventional question defines the core philosophy behind Prof. Roy’s work, which reshapes how machines operate in real-world scenarios, including flood relief, agriculture, and military logistics.
Textbook engineering produces predictable outcomes, with forces following equations. In practice, especially with aerial systems, conditions are far less controlled. Drones contend with wind, drag, shifting payloads, and numerous environmental disturbances that are difficult to model precisely.
Prof. Roy, who does not have a traditional mechanical background, embraces uncertainty rather than resisting it. His lab develops systems that adapt in real time with minimal prior knowledge and rely on fundamental physical principles instead of extensive datasets.
This philosophy appears in the lab’s work on drone dynamics. In a flood rescue scenario, for example, a drone delivers supplies onto a moving platform, tracks motion and compensates for disturbances with precision. The primary challenge emerges immediately after the drone releases the payload.
“When it drops, the drone suddenly shoots up, the dynamics switch completely,” Prof. Roy explains. In that moment, the system’s weight, balance, and governing rules change mid-air. The controller must anticipate and stabilize this switching dynamic, requiring adaptability rather than precision.
A similar challenge occurs when drones carry suspended loads. A swinging payload is not only inefficient but also poses safety risks, especially near people. “If a human is just under the payload, it’s pretty nasty,” Prof. Roy notes. The system must respond intelligently to disturbances and stabilize itself even as external forces affect its balance.
As the research progresses, drones take on roles beyond transportation. When equipped with a manipulator arm, a drone interacts with objects rather than simply moving them. However, each new component introduces instability and alters the center of gravity.
In one collaboration, the team developed a novel, motorless gripper inspired by the snap of a slap band. “The force must be enough to grasp but not so high that it destabilizes the drone,” Prof. Roy remarks. This situation presents an elegant, algorithmic challenge.
In another experiment, the lab addresses mid-air catching. Prof. Roy uses a cricket metaphor, “If the hand is hard, the ball drops. If it’s soft, it absorbs the impact.” This intuition led to software-driven compliance, enabling the robotic arm to absorb impact while the drone stays stable, mimicking the nuanced control of human movement.
Despite its technical ambition, the lab maintains an unhierarchical culture. Leadership shifts based on the task, students with hardware expertise lead system development, while Prof. Roy contributes to theory and control. This approach has produced nearly thirty journal publications in six years and ensures continuity, as students train their successors and ideas progress collectively.
The team has already implemented some of these ideas beyond the lab. A project on reconfigurable drones, which physically adapt to the shape and size of their payloads, led to a startup and applications with the armed forces. In these high-stakes situations, systems must function with minimal preparation, tight timelines, and uncertain conditions. These environments test the philosophy of adaptive control most directly, where failure has significant consequences.
Not all applications are high stakes, but they remain impactful. In a collaboration on drone-based pollination, the lab tackled agricultural challenges, where inefficiency is both common and costly. Traditional pollination requires labor-intensive and time-sensitive manual agitation of crops within a limited timeframe.
Initial drone-based pollination attempts failed. Prof. Roy’s team instead developed a system that combines airflow with gentle physical interaction to maximize pollen release without harming plants. The results exceeded collaborators’ expectations. Across multiple field trials, the drone-assisted method did not just match manual pollination, it surpassed it.
Looking ahead, Prof. Roy identifies new challenges, developing robotic hands that manipulate objects without costly sensors, systems that infer touch through motion, and machines that further bridge the gap between knowledge and action.
The unifying element is not a specific application, but a mindset. “You don’t need to know everything about a system to make it work,” he says. In a field that focuses on precision and prediction, this idea is subtle yet radical. The real world rarely provides complete information due to its variability and unpredictability. The challenge is not to eliminate uncertainty, but to design systems that can operate within it. Prof. Spandan Roy’s lab finds solutions in this space between control and chaos.
